MCP elicitation

What is MCP elicitation?

MCP elicitation is an MCP capability that lets a server ask the user a question through the client during a tool call, supporting interactive workflows. In practice, it gives agents a structured way to request missing input before continuing. (modelcontextprotocol.io)

Understanding MCP elicitation

MCP elicitation is designed for moments when a tool cannot safely or correctly proceed without more context. Instead of guessing, the server can request user input through the MCP client, then resume the workflow with that answer. The official MCP spec describes this as a server-initiated request for additional information from the user, nested inside other MCP server features. (modelcontextprotocol.io)

This matters most in agentic systems where tool calls are not one-shot operations. A retrieval flow might need the user to choose a project, a date range, or a permission scope before the server can finish the task. Elicitation keeps that interaction inside the protocol, which makes the handoff between model, server, and client clearer for builders working on interactive assistants.(modelcontextprotocol.io)

Key aspects of MCP elicitation include:

1. Server-initiated: the question originates from the MCP server, not from a separate side channel.
2. Client-mediated: the MCP client presents the prompt to the user and returns the response.
3. Workflow-aware: it fits inside an ongoing tool call instead of breaking the conversation flow.
4. Structured input: the request can capture the exact information the server needs.
5. Interactive by design: it supports human-in-the-loop experiences when automation needs clarification.

Advantages of MCP elicitation

1. Fewer failed tool calls: the server can ask for missing context before executing.
2. Better user experience: users answer a targeted question instead of restarting a workflow.
3. Cleaner agent design: interactive steps stay inside one protocol, which simplifies orchestration.
4. Safer execution: sensitive or ambiguous actions can pause for confirmation.
5. More flexible assistants: teams can build flows that adapt to what the user actually provides.

Challenges in MCP elicitation

1. Prompt design: the question must be precise enough to avoid confusing the user.
2. State handling: the server needs to resume correctly after the user responds.
3. Latency tradeoffs: every extra question adds another step in the workflow.
4. Client support: the experience depends on the MCP client implementing elicitation well.
5. Edge cases: teams need to think through cancellations, retries, and partial answers.

Example of MCP elicitation in action

Scenario: a support agent is trying to create a refund request, but the server needs the customer’s order number before it can continue.

The MCP server initiates elicitation through the client and asks the user for the order ID. Once the user responds, the tool call resumes with the new value and completes the refund lookup. The result is a smoother flow than failing the request and forcing the user to start over.

This pattern is especially useful for agents that gather inputs dynamically, such as onboarding assistants, internal ops tools, and approval workflows.

How PromptLayer helps with MCP elicitation

PromptLayer gives teams a place to manage the prompts, guardrails, and evaluation logic around interactive MCP workflows. That makes it easier to iterate on elicitation questions, review how users respond, and track whether the workflow is actually reducing friction.

Ready to try it yourself? Sign up for PromptLayer and start managing your prompts in minutes.